← back
How to build Enterprise Aware Agents - Chau Tran, Glean
Takeaway
Stop choosing between workflows and agents—train agents on your golden workflows and let agents mine new workflows from real user usage.
Summary
- Glean's Chau Tran reframes the workflow-vs-agent debate: an agent's trace IS a workflow, so the two are complementary—use workflows as evals and training data for agents.
- Workflows = predictable, low cost, easy to debug (Toyota); agents = open-ended, higher cost, occasional brilliance (Tesla).
- Three synergies: workflows as agent evals (judge intermediate steps not just outputs), workflows as agent training data via SFT/RLHF, and agents as workflow-discovery/generation engines.
- 'Enterprise-aware AGI' must learn company-specific protocols—closing the gap between an acceptable output and a great one for tasks like competitor analysis.
agentsworkflowsenterprise
Original description
While LLMs demonstrated impressive reasoning capabilities, their out-of-the-box reasoning is akin to hiring a brilliant but brand-new employee who doesn’t have the enterprise context of “how things are done at this company”. In this talk, I'll introduce “Workflow Search” as a paradigm to build enterprise-aware agents that can balance predictability on common tasks, and flexibility on unforeseen tasks. About Chau Tran Chau Tran is a Software Engineer at Glean, currently leading the technical work on Glean Assistant and semantic search. They have been with Glean for over 3 years and have a history of impactful contributions in engineering teams. Previously, Chau worked as a Research Engineer at FAIR within Meta and held technical roles at Quora. They graduated from Brown University with a Bachelor's degree in Computer Science. Recorded at the AI Engineer World's Fair in San Francisco. Stay up to date on our upcoming events and content by joining our newsletter here: https://www.ai.engineer/newsletter